This is a whirlwind tour of some basic cartographic design topics. For lots more information and practical resouces to go along with these, look at resources.md in this repository.
Silly question, right? No! Think about the what and why of maps.
A map is a representation of a place. It is a symbolic interpretation of place and highlights the relationships between elements in space, either perceived or actual. It reflects choices and biases of the mapmaker. It does not and cannot represent everything in the place. Things must be omitted, simplified, etc. for the map to make sense.
Maps can be divided into two broad categories: reference and thematic.
Reference maps emphasize the location of spatial phenomena, e.g. topographic maps or road maps.
Thematic maps emphasize the spatial pattern of geographic attributes, e.g. population density or income. Thematic maps come in a variety of flavors, including:
Choropleth maps in which areas are shaded by color to represent values. Choropleth maps are perhaps the most common and familiar type of thematic map, frequently seen for demographic or political data.
Proportional symbol maps, in which symbols (such a circles) are scaled in proportion to the data they represent. Symbols might represent point data, but it is also common for proportional symbols to be used for area features.
Dot maps, which use points to represent counts of a geographic phenomenon. Dot density maps rely on visual scatter to show spatial pattern. Dot maps may have a one-to-one relationship, in which one dot represents a single instance of something (e.g. one dot per vote), or a one-to-many relationship in which each dot represents a certain quantity of something.
Contour or isarithmic maps interpolate values between points. They're most recognized as useful maps for representation of elevation or depth, but they can also be used to represent more thematic phenomena, such as temperature, weather, or social phenomena.
...and many other types! These notes are mostly about general map design topics rather than specific types of maps, but see the resources list for more about some types of thematic maps.
It’s always good to ask whether your data or story even has a spatial component. If geography is not important, a map is probably not the best way to show it. Additionally, if you’re working on the web, think about how much your map benefits from interactivity, if at all. Just because you can add interaction and more data doesn’t mean you should—it requires a lot more design work from you, and potential difficulty for your audience.
Visual variables are the basic ways in which graphical marks vary. Some allow quick visual grouping of symbols: for example symbols are easily grouped by hue, but not shape. Some are perceived as naturally ordered, such as size and value (lightness). Does the data you’re mapping have order (or quanitity)? Are there groups that should stand out? Be mindful of the appropriate visual variables to use in your symbolization.
The seven “retinal variables” proposed by Jacques Bertin in Semiology of Graphics (1967). Later cartographic researchers have expanded the list.
Good visual hierarchy is crucial to overall effective map design. Visual hierarchy is the organization of design such that some things seem more prominent and important, and others less so. Visual hierarchy should match the intellectual hierarchy of your composition—what are the most important things in your map, conceptually? Those should stand out. Visual hiearchy depends on figure-ground relationships. Figures are the things that stand out; ground is the rest. (Think foreground and background.) Contrast is the key. In general, larger and darker things appear as figure.
Try the squint test. If you stand back or squint to blur your map, do the key components still stand out?
Color is an important component of good map design, and while true mastery of color takes a designer’s eye and experience, there are some general guidelines to start with:
- Keep in mind cartographic conventions you’ve seen: blue for water, etc.
- Avoid red–green color schemes, as a significant population is colorblind and can’t discern these.
- Think about whether your colors imply relationships, and whether relationships actually exist in what you’re mapping. If two things are not related, it may be best to color them with different hues.
- Subtlety is usually your friend! Save any bold colors for important things in your hierarchy.
- When using color ramps, try to limit yourself to 6-8 colors or shades. Beyond that, it becomes challenging for the human eye to differentiate.
For choropleth mapping, use color ramps appropriate to the nature of your data. If there is no natural order of your data classes, use a qualitative color scheme. If the data have order in one direction, use a sequential color scheme. (Definitely NOT a rainbow scheme!) If the data have order and can be divided above and below some meaningful breakpoint, use a diverging color scheme.
Most choropleth maps use a classification scheme to group data values into bins, ideally 3–7 of them. This can make it easier to read the map—but different classification methods can produce vastly different looking maps. They all have their advantages and drawbacks. Try to use one that is both understandable and clarifies real patterns in the data. A few common methods:
- Equal interval: divide data into groups of equal value ranges
- Quantiles: put the same number of data observations in each class
- Natural or Optimal breaks: find natural groupings, maximizing similarity within groups and differences between groups
Equal interval with four classes. No values at all in the third class!
Quantiles. Does that outlier at the end really belong with the other green dots?
Hypothetical natural breaks. Looks good but can be hard to understand.
Exact same data, different classification methods. Example by John Nelson.
Choropleth maps should also be normalized, that is, mapped as some kind of ratio rather than raw counts of things. Otherwise you just get http://xkcd.com/1138/ and furthermore your visual interpretation of values is affected by relative sizes of areas. For example, population density or GDP per capita are more meaningful than just population or total GDP.
Displaying a round earth on a flat map, i.e., projecting the map, requires distortions. Map projections can accurately maintain local angles (shape, more or less), sizes, directions, or distance—but never all of them. From a design perspective, think about a few things:
- Is the projection appropriate for the type of map and the map's purpose?
- What is the area covered by the map?
- Does the map projection look decent, with geography looking recognizable?
Choropleth and dot density maps should use equal-area projections, such as Albers Equal Area Conic (above) for continent-level maps or Mollweide for world maps, so that size distortions don't lead us to misinterpret data values.
Compromise projections such as Winkel Tripel (above) or Robinson were developed mainly to produce a familiar, not wildly distorted appearance. They preserve no properties perfectly but generally look good for world maps.
The widespread Mercator projection (seen in most web maps) preserves local directions but badly distorts sizes at global scale. It's good for navigating in a city or sailing across the ocea, but it is not ideal for thematic maps.
Lambert Conformal Conic is a popular projection for North America and other areas of similar east-west extent. It preserves shape in a sense and generally looks "right" and familiar.
Type can convey information through words, but characteristics of type can also communicate meaning. Type on maps is found not only in labels, but also in supplemental information, such as legends and sources, and in blocks of prose. Do not take typography for granted! Your choices of typeface, style, etc. can have a strong impact on the clarity, meaning, and tone of the map.
- Be consistent. A good rule of thumb is to choose two fonts: one serif font, and one sans-serif.
- Encode information in different weights and styles (bold, italic, regular, light, etc.).
- Use principles of your visual and intellectual hierarchy with text. Make important things bolder and bigger, and less important things smaller and lighter (or even, off the map!).
- A convention is to label physical features (water bodies, for example) with a serif font. Water labels are often italicized.
Hypothetical label styles and hierarchy. Labels for large areas often use increased letter spacing (tracking).
Masking or halos around labels can improve legibility against background features. (Map example by Daniel Huffman.)
Order of preference for label placement on point features. (source)
At its core, cartography is about abstraction. We don’t show data in its raw form; we clarify it in a variety of ways, often by removing things. A big reason is scale, that is, the size of the map compared to the size of the real world. It simply isn’t possible to show every tiny detail! Data and graphics should be generalized appropriately to the map scale: at a large scale (”zoomed in”) you can have more detail. Typical generalization operations include:
Selection: choosing which objects to include on the map
Simplification: reduce the number of vertices in an object
Smoothing: reduce sharp angles to smoother curves
Aggregation: group points into areas
Amalgamation: group areas into larger areas
Collapse: reduce a detailed object to a point symbol
Merge: grouping of line features
Refinement: select only portions of an object to display
Exaggeration: amplify a part of an object
Enhancement: add detail that visually elevates an object
Displacement: increase separation between objects
Many links to in-depth information and practical tools related to all these topics are in resources.md!